2026-09-09 · research · Dyplux
BNB Chain's activity has a new layer: measuring automation and adoption
27.2% of 131.97M canonical USDT transfers went wallet to wallet, while 48.6% moved contract to contract. A behavioural view of how direct users, automation and programmable settlement contribute to BNB Chain activity.
Why we asked
A transfer count describes activity, not one single kind of user. Routers, automated strategies, exchange hot wallets and direct wallet payments all contribute to the total; some automated flows may also support AI-enabled workflows. We wanted a more useful question for anyone studying adoption: how much movement is wallet-to-wallet, how much is contract-to-contract, and what can each layer tell us about a programmable network?
How we measured
We measured the canonical BNB Chain USDT contract (Tether USD, 0x55d398326f99059ff775485246999027b3197955) over the seven days ending 2026-09-09, using Dune's Spellbook tables. The contract identity was checked on-chain at the recorded block: name() = Tether USD, symbol() = USDT, and decimals() = 18. A ticker is metadata, not identity, so no headline number uses ticker matching alone.
An address is a contract if it appears in bnb.creation_traces. Everything else is counted as an EOA — a wallet. “Wallet to wallet” therefore means EOA to EOA. It is a statement about code, not about identity.
Three public queries make up the measurement: Q1a splits transfers by sender and recipient type; Q2 measures frequency and fan-out inside wallet-to-wallet transfers; Q3 measures the amount distribution. Q2 and Q3 independently total 35,925,008 transfers, which is the cross-check.
Dune: 8653665 · 8653636 · 8653649
Almost half of the count runs through contracts
Of 131,971,472 canonical USDT transfers, 27.2% were wallet-to-wallet. Contract-to-contract transfers were the largest bucket by count, produced by 52,834 distinct senders. This is a view into the programmable layer of BNB Chain: smart-contract routes, exchange operations and automated services working together.
| Path | Transfers | Share | Senders | Value |
|---|---|---|---|---|
| Wallet → wallet | 35,958,380 | 27.2% | 6,517,498 | $17.31B |
| Wallet → contract | 14,946,612 | 11.3% | 1,065,618 | $11.71B |
| Contract → wallet | 16,982,084 | 12.9% | 61,784 | $11.68B |
| Contract → contract | 64,084,396 | 48.6% | 52,834 | $17.64B |
A focused automated layer produces a meaningful share of wallet activity
Within the wallet-to-wallet slice, 722 addresses sent more than 1,000 times in the week and produced 26.9% of transfers. Another 7,905 sent to more than 50 distinct recipients and produced 30.9%. Those are behavioural signatures compatible with scripts, distributors, exchange hot wallets, airdroppers and other automated services — not a count of identified bots.
| Transfers per sender | Senders | Transfers | Share |
|---|---|---|---|
| 1 | 3,265,031 | 3,265,031 | 9.1% |
| 2–5 | 2,313,624 | 6,409,260 | 17.8% |
| 6–20 | 746,646 | 7,529,300 | 21.0% |
| 21–100 | 173,366 | 6,219,775 | 17.3% |
| 101–1,000 | 13,456 | 2,822,501 | 7.9% |
| >1,000 | 722 | 9,679,141 | 26.9% |
| Distinct recipients | Senders | Transfers | Share |
|---|---|---|---|
| 1 | 4,614,494 | 10,379,669 | 28.9% |
| 2–5 | 1,682,102 | 9,383,966 | 26.1% |
| 6–50 | 208,344 | 5,060,645 | 14.1% |
| >50 | 7,905 | 11,100,728 | 30.9% |
Most wallet-to-wallet transfers are worth less than a cent
22,278,969 wallet-to-wallet transfers, 62.0% of the total, were worth less than $0.01. Together they represented $1,511. This is consistent with address poisoning, distribution campaigns and other low-value automated flows, but the query measures amounts, not intent.
| Amount | Transfers | Share | Total value |
|---|---|---|---|
| < $0.01 | 22,278,969 | 62.0% | $1,511 |
| $0.01–$1 | 812,131 | 2.3% | $195,641 |
| $1–$10 | 2,057,969 | 5.7% | $10.4M |
| $10–$100 | 5,267,489 | 14.7% | $205.3M |
| $100–$1,000 | 4,255,419 | 11.8% | $1.28B |
| $1,000–$10,000 | 1,087,410 | 3.0% | $3.01B |
| > $10,000 | 165,621 | 0.46% | $12.78B |
Remove the sub-cent transfers and 13,646,039 remain, carrying $17.29B. The saved buckets also show that 64.28% were below $1; that is an arithmetic derivation from the two sub-$1 rows, not a separate dust label.
What this measurement leaves open
- An EOA is not a person. It can be a human, a script or an exchange hot wallet.
- Exchanges were not separated: the label join timed out twice on Dune's free execution limit.
- The frequency and amount distributions cannot be crossed; we cannot say what share is both dust and produced by the 722 high-frequency senders.
- EIP-7702 delegated accounts count as EOAs in this method because they do not appear in
creation_traces. - This is one token, one seven-day window and one chain — a reading, not a trend.
What we measure next
Repeat the measurement weekly, then run the same grid for Ethereum and Base, test the exchange-label split on a shorter window, and measure EIP-7702 delegated accounts through authorization_list. The useful output is a reusable adoption view: direct wallet activity alongside the programmable layer that supports exchange infrastructure, stablecoin flows, automated strategies and future agentic workflows.
Reproduce it
The three Dune queries are public above. The method, the SQL and the raw dated receipts are kept with the research record; the contract identity receipt is linked from the source record rather than hidden behind a model or a live API call. The site itself is a static publication of this dated snapshot.
Open the public receipt manifest →
Dyplux is an independent research desk in Lisbon. The author works at CoinMarketCap; this piece was not reviewed by CoinMarketCap or Binance, and it is not financial advice.